Improving Fairness and Privacy in Selection Problems

Supervised learning models have been increasingly used for making decisions about individuals in applications such as hiring, lending, and college admission. These may inherit pre-existing biases from training datasets discriminate against protected attributes (e.g., race or gender). In addition to unfairness, privacy concerns also arise when the use of reveals sensitive personal information. Among various notions, differential has become popular recent years. this work, we study possibility using a differentially private exponential mechanism post-processing step improve both fairness supervised models. Unlike many existing works, consider scenario where model is select limited number applicants available positions limited. This assumption well-suited scenarios, job application We ``equal opportunity'' notion show that mechanisms can make decision-making process perfectly fair. Moreover, experiments on real-world fairness, with slight decrease accuracy compared without post-processing.